THEJOURNALOF BIOLOGICALCHEMISTRY

Val. 258?No. 22, Issue of November 25, pp. 13825-13832,1983

Printed in U.S.A.

The Pathway of ATP Hydrolysis by Dynein

KINETICS OF A PRESTEADYSTATEPHOSPHATEBURST*
(Received for publication, April 1, 1983)

Kenneth A. Johnson
From the Department of Biochemistry, Microbiology, Molecular and Cell Biology, The Pennsylvania State University,
University Park, Pennsylvania 16802

~~~~~

* This work was supported by National Institutes of Health Grant

GM-26726. The costs of publication of this article were defrayed in
part by the payment of page charges. This article must therefore be
hereby marked advertisement in accordance with 18 U.S.C. Section
1734 solely to indicate this fact.

We have failed to observe a phosphate burst in isolated 14 S dynein,

but further work is currently underway using higher concentrations
of protein and a newly constructed device able to use smaller sample
volumes. The concentrationof 30 S dynein was calculated as 80% of
the total protein. The calculation of the molecular weight per ATP
site (see Fig. 2) is therefore based upon the assumption that the 14 S
dynein does not contribute to theobserved signal.

13825

Downloaded from www.jbc.org by guest, on December 25, 2012

The kinetics of ATP binding and hydrolysis (forma- traction (Huxley, 1969; Taylor, 1979). Without committing
tion of acid-labile phosphate) by the Tetrahymena 30 oneself to theprecise conformational statesof the dynein arm
S dynein ATPase has beenmeasuredbychemical
in each step of the cycle, this model provides a framework
quench flow methods. The amplitudeof the ATP-bind- with which to examine the reactions involving the coupling
ing transient gave a molecular weight per ATP-bind- of ATP binding and hydrolysis to the association and dissoing site of approximately 750,000, suggesting nearly ciation of dynein with theB-subfiber.
3 ATPbindingsites12million
M , dynein molecule
Since ATP causes the release of the rigor bond, the first
(Johnson, K. A., and Wall,J. S . (1983) J. Cell Biol. 96, question to address is whether ATP hydrolysis precedes or
669-678). ATP binding occurred at
the rate predicted follows the ATP-induced dissociation. Our previous kinetic
from the apparent second orderrate constant of 4.7 X analysis showed that ATP induces a n extremely rapid dissolo6 M s measured by analysis of the ATP-induced ciation of dynein from the microtubule. Stopped flow lightdissociation of the microtubule-dynein complex (Porter, M. E., and Johnson, K.A. (1983) J. Biol. Chem. scattering measurements established that thelifetime of the
258,6582-6587). Hydrolysis was slower than binding microtubule-dynein-nucleotide complex following ATP bindand occurred ata rate of 55 s-, at 30 and 50 PM ATP. ing was much less that 1 ms. The apparent second order rate
The rate limiting step for steady state turnover (prod- constant for ATP bindingwas 4.7 x lo6 M- s-.
T o establishthepathway
by which ATP hydrolysis is
uct release) occurred with a rate constant of 8 s.
These data show that first
the two steps of the pathway coupled to the microtubule-dynein cross-bridge cycle, it is
of coupling ATPhydrolysis to the microtubule-dynein necessary to measure the kinetics of ATP binding and hycross-bridge cycle are the same as those described by drolysis on a time scale short enough to resolve the transients.
Previous analysis of the dynein ATPase has indicated the
Lymn and Taylor for actomyosin (Lymn, R. W., and
Taylor, E. W. (1971) Biochemistry 10, 4617-4624).
existence of a presteady state phosphate burst, based upon
Namely, ATP binding induces the
very rapid dissocia- extrapolation of steady state data atlow temperature (Shimtion of dynein from the microtubule and ATP hydrol- izu et al., 1979; Takahashi and Tonomura,1979; Terashita et
ysis occurs more slowly following dissociation. More- al., 1983) or relatively short times (Evans, 1982). These data
over, in spite of rathergross structural differences, the suggest the presence of an enzyme-bound product intermedikinetic constants for dynein and myosinare quite sim- ate with a long lifetime relative to its rate of formation and
ilar.
have been used to argue that dyneinmay share some kinetic
properties with myosin.
In this report, I measure the kinetics of ATP binding and
Dynein, as originally defined (Gibbons andRowe, 1965), is hydrolysis by chemical quench flow methods and show that
a general name describing any
high molecular weight ATPase hydrolysis is slower than the ATP binding-induced dissociathat interacts with microtubules (Gibbons, 1981). The most tion of the dyneinfrom the microtubule. Preliminary reports
extensively studied is the outer arm dynein isolatedcilia
from of these datahave been presented (Johnson and Porter,1981,
1982a, 1982b).
and flagella. Numerousreportshaveestablishedthatthis
dynein is permanently attached to the A-tubuleof the outer
MATERIALSANDMETHODS
doublet and interacts transiently with the adjacent B-subfiber
Preparation of Proteins-Dynein, tubulin, and the microtubuleto produce a force for sliding (Summers and Gibbons, 1971; dynein complex were prepared as described previously (Porter and
Johnson, 1983a). Column-purified dynein, obtained by chromatograSatir, 1968; for reviews, see Gibbons, 1981, or Warner and
Mitchell, 1980). The interactionof dynein with the
B-subfiber phy on DEAE-Sephacel (Pharmacia Fine Chemicls, Piscataway, NJ),
is coupled to the binding and hydrolysis of ATP (Brokaw, was used without further purification. This was dictated by the
of material needed for the chemical quench flow experi1975; Brokaw and Benedict,
1968; Brokaw andGibbons, 1975; quantities
ments. The column-purified dynein consists of a mixture of 30 S
Gibbons and Gibbons, 1972). A rigor-type bond is formed in dynein (80%)and 14 S dynein (15%).Since the chemical quench flow
the absence of ATP and released by the addition of ATP experiment provides a signal that is directly proportionalto themolar
(Gibbons and Gibbons, 1974; Gibbons, 1975; Zanetti et al., concentration of enzyme sites, the contribution of the 14 S dynein to
1979). Therefore it is reasonable to propose a simple, four- the observed signal should be less than or equal to 15%, based upon
state cross-bridge cycle analogous to models for muscle con- our preliminary estimate of the molecular weight of the 14 S dynein.

13826

State
Presteady
Dynein

TIME

FIG. 1. Chemical quench flow apparatus. A, a schematic representation of the quench flow apparatus is presented where the
squures represent mixers with the delay line (reaction chamber) in
between. The arrows show the positions of lines used to flush the
expended sample from the device between consecutive runs. B, the
position of the pistons, in arbitrary units, is shown as a function of
time (20 ms/division). Note that the pistons start abruptly and
continue linearly until they stop. The arrows indicate the start and
stop times.

ical quench flow apparatus. Dynein and [y3'P]ATP were manually

mixed and thereaction was stopped with 1 N perchloric acid to obtain
reaction times from 3-30 s.
The concentration of [32P]phosphatewas determined by applying
the sample to a charcoal column to adsorb unreacted[y3*P]ATP and
measuring the radioactivity of the effluent relative to thetotal sample
as described by Bagshaw and Trentham (1973) and Taylor (1977),
with the following changes. For each samplea 2-ml column was
prepared using activated coconut charcoal (50-200 mesh, Fisher
Scientific Co., Pittsburgh, PA); this charcoal provided both a rapid
flow rate and quantitative
retention of [y"P]ATP while greater than
95% of the phosphate was recovered in the eluate. The charcoal was
extensively washed before use with a solution consisting of 65 mM
HnPO, and 10 mM Nap207 (wash solution). The wash solution and
columns were kept cold (0-4 "C). Immediately after stopping the
reaction with acid, 0.5mlof
the wash solution was added to the
sample and then a 0.8-ml aliquot of the mixture was placed onto a
charcoal column. The column was then washed three times with 2.5
ml of the wash solution and theeffluent was collected directly into a
plastic scintillation vial (sample A). A 0.2-ml aliquot of the reaction
' D. B. Clutter and K. A. Johnson, manuscript in preparation.
mixture was added directly to another scintillation vial containing 8
'The abbreviations used are: PIPES, piperazine-N,N'-bis(2-eth- ml of wash solution (sample B).The radioactivity of each was
anesulfonic acid); STEM, scanning transmissionelectron microscopy. determined by the Cerenkov method using a Beckman LS 7000 liquid
K. A. Johnson, manuscript in preparation.
scintillation counter (Beckman Instruments, Inc., Irvine, CA) on the
' K. M. Arndt and K. A. Johnson, unpublished observations.
tritium channel. The concentration of phosphate was then calculated

Downloaded from www.jbc.org by guest, on December 25, 2012

The concentration of dynein was determined by readings of absorbance a t 280 nm using an extinction coefficient of 0.97 cm'/mg
based upon a calibration using dry weight of 30 S dynein.' This was
within 3% of the value obtained by the Lowry method. The concentration of tubulin was determined by the Lowry method, with a
bovine serum albumin standard.
All of the experiments reported here were performed a t 28 "C in a
buffer consisting of 50 mM PIPES' and 4 mMMgCIZ titrated to pH
7.0 with NaOH. The experiments were completed within 24 h of the
preparation of the protein.
Stopped Flow Light-scattering Measurements-Light scattering a t
90" to the incident beam was measured as described by Porter and
Johnson (1983b) using a stopped flow spectrofluorimeter. Data were
collected and stored digitally using a microcomputer (On-Line-Instruments Systems, Inc., Jefferson, GA). The datawere fit to a single
exponential by a modification of the Method of Moments (Dyson and
Isenberg, 1971).3
The microtubule-dynein complex was formed a t -1 mg/ml of
tubulin and dynein, as described previously (Porter and Johnson,
1983a, 1983b3, then diluted 5-fold with warm buffer and immediately
loaded into one syringe of the stopped flow apparatus. The reaction
was initiated by mixing the microtubule-dynein complex 1:l with
ATP in the same buffer. All of the ATP concentrations reported are
the concentrations after mixing.
Chemical QuenchFlow Measurements-The kinetics of ATP binding and hydrolysis was measured over the period of 5-100 ms using a
quench flow apparatus constructed in this laboratory. The device
consisted of three syringes, as diagrammed in Fig. lA, which were
simultaneously driven by an aircylinder to force the mixing of dynein
first with substrate, [y-"PIATP, and then with a quench solution.
The solutions were mixed using Berger ball-type mixers (Research
Instruments andMfg. Co., San Diego, CA). Each time pointrequired
200 pl of enzyme, 200 pl of [y-"PIATP, and 400 pl of quench solution.
Sample lines were flushed between consecutive runs as indicated in
Fig. 1A. The time between mixing with substrate and the quench
solution was varied from 5-100 ms by varying the length of the delay
line between the two mixers and the rate atwhich the syringes were
driven. The volume of solution between the mixers (delay volume)
was measured using ["P]phosphate as a tracerby loading and flushing
in the appropriate sequence to recover the solution that was contained
between the two mixers. The progress of the pistons was monitored
using a linear position transducer (Bourns, Riverside, CA) connected
in series with a constant voltage supply, the outputof which was fed
into a storage oscilloscope. As shown in Fig. lB, themovement of the
pistons was linear with time. The progress of the pistons was timed
(drive time = time from start to stop) and used to calculate the
reaction time for each sample collected according to the equation:
reaction time = drive time X (delay volume/sample volume), where
the sample volume was the sum of the enzyme and substratevolumes
(400 p l ) .
The time course of ATP hydrolysis (formation of acid-labile phosphate) was obtained by quenching with 1 N perchloric acid (final
concentration) in the chemical quench flow apparatusandthen
measuring the concentration of phosphate released in each sample,
as described below.
The time course of ATP binding was measured by quenching with
a 200-fold excess of unlabeled ATP, allowing the reaction to continue
for 1-2 s (10-20 half-lives of the enzyme turnover) and thenstopping
the reaction by manually adding 1 N perchloric acid. In this millisecond pulse-chase experiment, any tightly bound [y3'P]ATP will be
recovered as ["P]phosphate. The term tightly bound is operationally
defined with the requirement that the rate of the forward reaction
(hydrolysis and product release) must be 1 order of magnitude or
more faster than the rateof substrate dissociation to get quantitative
recovery of the bound ATP as phosphate. The observed increase in
recovery of phosphate in the cold chase experiment provides evidence
in supportof a relatively slow substrate dissociation step. Preliminary
evidence based upon measurement of the rate of synthesis of ATP
from ADP and phosphate supports this assumption.'
The steady state turnover rate was determined immediately following each quench flow experiment using the same dynein and ATP
samples and under conditions identical with those used in the chem-

Phosphate Burst

Burst
Phosphate
State
Presteady
Dynein

13827
days of synthesis to ensurea background of [32P]phosphate lessthan
1% relative to [Y-~*P]ATP. In the
two experiments shown in Fig. 2,
A and B, backgrounds of 0.74 and 0.80% were recorded, corresponding
to blanks of 0.24 and 0.37 p ~ respectively.
,
Although these blanks
were relatively large, they were constant within 4-6% in replicate
measurements. The blanks were subtracted from the data to obtain
the concentration of phosphate produced during enzymatic hydrolysis. These datawere then fitted asdescribed in the Appendix to obtain
a measurement of the number of phosphates produced per enzyme
site. The concentration of phosphate was used to calculate the concentration of ATPase sites and thus themolecular weight per site.
The concentration of ATP in each experiment was determined by
absorbance measurements usingan extinction coefficient of 15.0 cm/
*mol. A trace of labeled ATP was then added to obtain anactivity of
approximately 500,000 cpm/ml of sample giving a specific activity of
-0.01 Ci/mmol of ATP.

1.5

1.2

0.9

0.6

RESULTS

0.3

TIME (msec)

v)

k
kz
+ ATP 1
E-ATP a E-ADP-P,

k3

k-,

+ ADP + Pi

where E represents a dynein enzymatic site. The data were

fit to the complete solution of the above equation using all
I
four rate constants asdescribed in theAppendix. The smooth
a
u,
lines in Fig. 2 were computed according to Equations 9 and
0
10 (see Appendix) withthe rate constantsgiven in the legend.
I
Table I summarizes the best estimate of the rate constants
n
consistent with both setsof data.
The data shown were obtained using two separate enzyme
preparations and are representative of six separate experiments.Identicalresults
were obtained using a preformed
20
40
60
80
IO0
microtubule-dynein complex in quench flow experiments a t
30 ~ L MATP(datanotshown).Theability
of oneset of
TIME (msec)
constants to quantitatively accountfor data obtained at two
FIG. 2. Kinetics of ATP binding and hydrolysis. The time ATP concentrations and using two separate enzyme prepacourses of the presteady state ATP binding (0)and hydrolysis (0) rationsincreases ones confidence in the accuracy of the
transients are shown for two different ATP concentrations. In each,
estimates. A discussion of the fitting method and an estimate
the smooth lines are defined by Equations 9 and 10 of the Appendix,
using the rate constantsgiven below. A , 30 p~ ATP. The concentra- of the probable error is described in t.he Appendix. Approxition of 30 S dynein was 0.27 mg/ml. The fitting process (see Mate- mate 90% confidence limits for the best estimatesof the rate
rials and Methods and Appendix) provided a concentration of sites constants in fitting each data set aregiven in Fig. 2.
of 0.34 p~ giving an apparent molecular weight per site of 790,000.
In each experiment, the rate of ATP binding was faster
The best estimates and 90% confidence limits for the rate constants
than
the observed rate of ATP hydrolysis. The pseudo-first
are: kl[ATP] = 120 k 30 s-, k2 = 55 & 12 s-, k-, = 12 & 9 s-, and
k, = 8 -+ 3 s-. B, 50 p~ ATP. The concentrationof 30 S dynein was order rate constant used to calculate the line for the ATPbinding transient (kl[ATP]) was 120 s at 30 /.LM ATP and
0.42 mg/ml. The fitting process provided a molar concentration of

sites of 0.56 p M giving an apparent molecular weight per site of

750,000. The best estimates and 90% confidence limits for the rate
constants are: kl[ATP] = 240 (>200) s, k2 = 52 & 9 s-, k-z = 10 f
6 s-l, and k3 = 9 & 3 s-.
as the ATP concentration of the original sample multiplied by the
activity of sample A divided by four times the activity of sample B.
[ Y - ~ ~ P ] Awas
T P synthesized by the method of Schendel and Wells
(1973) and purified on a 2-ml DEAE-Sephadex A-25 column (Pharmacia Fine Chemicals, Piscataway, NJ) eluted with steps of 0.1, 0.2,
and 0.4 M triethylammonium bicarbonate. The peak fractions of ATP
in the 0.4 M step were evaporated to dryness, resuspended in water,
and stored frozen until use. The [y3P]ATP was used within a few

TABLEI
Kinetic constants for the dynein ATPase
50 mM PIPES, 4 mM MgCI,, pH 7.0, 28 c.
Reaction

ATP binding
ATP hydrolysis
ATP synthesis
Burst amplitude
Product release
Michaelis constant

Kinetic
uarameter

4.7 kl
k?
k-z
ka
K m

Value

lo6 M
55 s-l
10 s-l
O.G/site
8 s
1.2 pM

Downloaded from www.jbc.org by guest, on December 25, 2012

The kinetics of ATP binding and hydrolysis was examined

in the millisecond time scale by chemical quench flow methods. Fig. 2, A and B, shows the results obtained at30 and 50
~ L ATP,
M
respectively (concentration during the reaction period) for two separate experiments using two different dynein
preparations. The data are plotted as the phosphate
released
per enzyme site, calculated by the fittingprocess described in
the Appendix. The filled symbols show the kinetics of ATP
hydrolysis (formation of acid-labile phosphate) and theopen
symbols show the kinetics of ATP binding, each of which is
followed by steady state turnover.
Thedatacanbeunderstoodinterms
of the following
pathway with the minimum numberof steps:

Dynein Presteady State Phosphate Burst

13828

was 240 s-' at 50 p~ ATP. These values are in agreement

with the apparentsecond order rate constantof 4.7 X lo6 M"
s-' (Porter and Johnson,198313). Although the data at 50 p~
ATP do not cover a short enough time interval to resolve the
kinetics of ATP binding, they demonstrate that the ratewas
at least faster than 200 s" and probably equal to the rate of
240 s-' obtained by stopped flow light-scattering measurements under the same conditions (see below).
The ATPhydrolysis transient was slowerthan theobserved
rate of ATP binding at each ATP concentration and was best
fit by forward and back rate constants of k2 = 55 s" and k-,
= 12 s-', respectively. A lag phase is apparent in the fitted
curve that is not necessarily demonstrated by the data obtained in the acid quench experiment. Rather, the fitting
process includes a lag phase since the kinetics of binding is
known and is slow enough to lead to a lag in the hydrolysis
reaction according to the complete solution to the transient

MD

k'
+ ATP 2
MD-ATP

I&
D

k
kz
k,
+ ATP A
D-ATP a D-ADP-Pi + D + ADP + P;
k-2

3.601
0

10

4.35

I
20

30

40

TIME (msec)

3 601

IO
TIME (msec)

15

20

FIG. 3. Kinetics of ATP-induced dissociation of the microtubule-dynein complex. The time course of the light-scattering
change, in arbitrary intensity units, is shown following the addition
of 30 PM ATP ( A ) or 50 PM ATP ( B ) to a preformed microtubuledynein complex in the stopped flow apparatus. The smooth lines are
the best fit of the data to a single exponential, with rates of 140 and
240 s-I, respectively.

where D represents dynein and M D represents the microtubule-dynein complex. The present results show that k, = k ' ]
and establish the values of the rate constants as given in
Table I. Data from Porter and Johnson (1983b) put a lower
limit on the rate of dissociation, k d 2 1000 s-'.
DISCUSSION

The observation and measurement of a presteady state

phosphate burst provide several important bits of information. The existence of a burst implies that some step in the
pathway following ATP hydrolysis is rate limiting during the
steady state turnover. Presumably, the rate-limiting step is
the release of products from the dynein-ADP-Pi intermediate.
Measurement of the amplitude of the ATP-binding transient
provides an active site titrationgiving the apparentmolecular
weight per ATP-binding site. The observed site M , = 750,000
gives 2.7 ATP-binding sites/2 million M,particle (Johnson
and Wall, 1983). Titration of the light-scattering amplitude
for ATP-induced dissociation of the microtubule-dynein complex also gave an apparentmolecular weight per ATP-binding
site of 750,000 (Shimizu and Johnson, 1983b). Allowing for
10% inactive protein, the most reasonable interpretation of
available data would suggest that there is one ATP-binding
site on each of the three dynein heads (Johnson and Wall,
1983), but this remains to be firmly established.
It is interesting to note that burst measurements based
upon extrapolation of steady state datafor sea urchin dynein
has given an amplitude of 1 ATP site/650,000 Da (Evans,
1982). Preliminary STEM analysis of sea urchin dynein-1has
indicated a two-headed particle with M , = 1.3 million: again
K. A. Johnson, I. R. Gibbons, and J. S. Wall, unpublished observations.

Downloaded from www.jbc.org by guest, on December 25, 2012

(see Appendix). Omission of the lag from the fitting process

would lead to anunderestimate in the rate of hydrolysis.
There is considerable margin for error in the estimate of
the rate of the back reaction, k-,, since it is dependent only
upon the relative amplitude of the ATP hydrolysis transient
and a %fold variationin rate would lead to only a 20%
variation in amplitude. Preliminary estimates of k-* by analysis of medium ['80]phosphate-H20 exchange (Hackney et
al., 1980), catalyzed by dynein in the presence of ADP, has
given a value of 4 s" for the rate of ATP ~ynthesis.~
The slope of the curves following the transient defines the
rate constant for product release equal to 8 s-I. This leads to
asteady state turnover of 6 s" since only -70% of the
enzyme is in the form of the enzyme product intermediate, as
defined by the values of the constants obtained for ATP
binding and hydrolysis (see Appendix, term ao).
The rate of ATP-induced dissociation of the microtubuledynein complex was measured by stopped flow light-scattering
methods, under conditionsidentical with those for the quench
flow experiments. The results obtained at 30 and 50 PM ATP
are shown in Fig. 3. The rateof the ATP-induced dissociation
at 30 p~ ATP (Fig. 3A) was 140 s" and the rate at 50 FM
ATP (Fig. 3B) was 240 s-'. These results show that the rate
of ATP-induced dissociation of the microtubule-dynein complex is equal to the observed rate of ATP binding, obtained
by chemical quench flow methods (Fig. 2) and thereby indicate
the following pathway:

Dynein
Burst
Phosphate
State
Presteady

13829

to the case for actomyosin (Taylor, 1979). In addition, the

microtubule-dynein complex could be dissociated
by ADP plus
vanadate;thisreaction
may representthe reverse of the
normal forward pathway involving recombination of the dyf
nein-ADP-Pi intermediate with the
microtubule. Attempts to
ADP + P
directly
establish
this
pathway
have
not been successful.
M*D
However, measurements of ATP turnover in reactivated
flagella (Brokaw, 1975; Brokaw and Benedict, 1968; Gibbons and
Gibbons, 1972) have indicated thattwo molecules of ATP are
hydrolyzed per dynein arm per beat of the flagellum. This
would require that the slowest step in the pathway be no
slower than the beatfrequency of 35 s-'. Product release from
free dynein, which occurs at a rate of 8 s-', is too slow to be
part of the cross-bridge cycle in vivoand one would predict
that microtubules should activate the steady state ATPase.
There has been no convincing demonstration of activation
of the dynein ATPase in solution
by microtubules. Following
ATP-induced dissociation, the dynein reassociates with the
D.ATP + M
D.ADP.P M
microtubule aftera lag period,
the durationof which is directly
FIG. 4. The pathway of ATP hydrolysis. A hypothetical cross- proportional to the ATP concentration (Porter,1982). Howbridge cycle is illustrated coupled to the ATPase cycle established
thus far, where M represents a microtubule and D represents dynein. ever, in experimentsto date, the rateof reassociation was not
It is suggested, but not yet established, that the D-ADP-Pi interme- fast enough to distinguish whether dynein reassociated with
diate (in parentheses) rebinds to the microtubule to complete the the microtubule before or after the release of products. Our
cycle. Each cycle results in the net upward movement of the adjacent preliminary measurementsof the rateof association of dynein
microtubule and thehydrolysis of one ATP.
with microtubules gave an apparent second order rate constant of 2 X IO4M" s-' (Porter, 1982). At tubulin concentrations that are accessible in solution (up to 100 pM), the rate
suggesting that there maybe one ATP binding site per
head. of dynein-microtubule association is too slow relative to the
The basis for the difference in the number of heads is cur- turnover rate of 6 s" to have a noticeable effect on the rate
of product release. However, the local concentration of tubulin
rently being studied.
The rate of ATP binding equals the rate of ATP-induced in the axoneme would be sufficiently high to account for the
dissociation of the microtubule-dynein complex. This obser- more rapid turnover in uivo. In fact, Gibbons and Brokaw
vation serves to define the firsttwo steps of the pathway by have observed a 4-fold increase in ATPase activityoccurring
which ATP hydrolysis is coupled to the cross-bridge interac- with motility (Brokaw and Benedict,1968; Gibbons and Gibwhich
tion as shown in Fig. 4. ATP binding causes a very rapid bons, 1972). This is consistent with our measurements
dissociation of dynein from the microtubule and hydrolysis might suggest an activation of -35:6 from the ratio of the
occurs as a slower step following dissociation. It is significant rate of turnover in the axoneme to the rate on the isolated
that thelifetime of the microtubule-dynein-ATP ternary
com- dynein molecule.
It has been suggested that calmodulin activates the steady
plex is muchless than 1ms (Porter and Johnson,
1983b). The
predominant pathway is the kinetically favorable route and state dynein ATPase ina calcium-sensitive manner (Blum et
although ATP hydrolysisconceivably could occur without al., 1980). However, it is likely that the release of products
dissociation (Stein et al., 1981), the lifetime of the ternary from the free dynein molecule, which determines the rate of
complex is too short to allow time for the reaction to occur steady state turnover, is not part of the cross-bridge cycle in
prior todissociation. The rapiddissociation is also significant vivo (see Fig. 4).Clearly, further work will berequired to
since therigor complex formed at the endof the power stroke establish whether calmodulin serves to regulate the microtumust have a short lifetime so that it will not impede the bule-dynein cross-bridge cycle and this cannot be addressed
slidingforces generated by other cross-bridges.Models of by studies on isolated dynein in the absenceof microtubules.
There are several models in the literature proposing conflagellar motility developed by Brokaw (1975) require this
occur to produce the force for sliding
rapid dissociation(>2,000 s-') at the endof the power stroke. formational changes that
The Michaelis constant for the isolated dynein, calculated (Goodenough and Heuser, 1982; Satir et al., 1981; Witman
and Minervini, 1982; Tsukita et al., 1983).However,all of
from ourpresteadystatedata(TableI),
agrees withthe
observed K,,, of 1 PM based upon steady state measurements these models are based upon interpretations of electron mi(Shimizu, 1981). In contrast, the ATP concentration depend-croscopic images of dynein in terms of a rather rigid, bulky
multisubunit arm. Our recent
STEM analysis of dyneins from
ence of the flagellar beat frequency gives an apparent Michaelis constant of-0.2
mM (Brokaw and Benedict, 1968; Tetrahymena (Johnson and Wall, 1983) and Chlamydomonas
Gibbons and Gibbons, 1972). This relativelyhighvalue
is (Witman et al., 1983) has indicated that theciliary outer arm
probably due toa more rapid rateof product release when the dynein consistsof a bouquet of three globular heads connected
ATPase is coupled to movement, but may also be due to the to the A-tubuleby three independent stems (see also Witman
requirement for a coordination of the interaction of many and Minervini, 1982). Our data indicate that threemolecules
dynein arms to produce thesliding velocity needed for wave of ATP are required dissociate
to
each dyneinmolecule (Shimpropagation.
izu and Johnson, 1983b) indicating that the three globular
Further work will be needed to establish the steps
involved heads interact with the B-subfiber to produce the force for
in recombinationof the dynein with the
microtubule. It should sliding. The simplest model would propose a rotation of the
be noted that analysis of the mechanism of vanadate inhibi- head on the microtubule surface to produce the sliding force
tion of dynein (Shimizu and Johnson, 1983a) suggests that analogous to current models for muscle contraction (Huxley,
phosphate release precedes ADP release and that therelease 1969; Taylor, 1979). The existing data can be interpreted in
change similar terms of this model.
of phosphate mayoccur with a large free energy

83.83

kATP

63.83

Dffl
+

Downloaded from www.jbc.org by guest, on December 25, 2012

Burst
Phosphate
State
Presteady
Dynein

13830

The Pathway of A T P Hydrolysis by Dynein: Derivation of the

Kinetics of A T P Binding and Hydrolysis
Typically, presteady state transient chemical quench flow
data are fitted aassingleexponential followed by linear steady
state turnover. This approach is valid only when substrate
binding and catalysis arewell separated or cannot be distinguished kinetically. However, if one wishes to simultaneously
resolve the kineticsof binding and catalysis
over a single time
interval, the experiments must
be performed under conditions
where the two rates differ by less than 5-fold. For example,
experimentsonthedyneinATPase
(see maintext) have
provided measurements of the kinetics of ATP binding and
hydrolysis where the two rates differed only by a factor of 3
or 5 . Under these conditions, oneexpects a lag in the formation of product due tothekinetics
of substrate binding.
Althoughchemical
quench flow dataare rarely accurate
enough to establish a lag, omission of the lag from the fitting
processleads to significant underestimation of the rate of
product formation. In this section, a derivation of the complete equationsdescribing binding andhydrolysis of ATP and
release of product is presented. Theuse of these equations in
fitting chemical quench flow data, using a computer graphics
subroutine, is then described.
The derivation was begun with the following model:
k2

d[E]/dt= -kl[E][S]+ k3[EP]

d[ES]/dt= k,[E][S]- kz[ES]+ k-z[EP]

(2)

d[EP]/dt = kP[ES]- Lz[EP] - k,[EP]

d[P]/dt= k,[EP]

Simultaneous solution of the equations gives the time dependence of each intermediate during the transient
where [E0]
= [ E ] [ES] [EP].

The release of phosphate with time is obtainedby integra= Jd[P]/[Eo]

= Jk3([EP]/[Eo])dt,
tion of the equation,[P]/[Eo]
substituting the time
dependence of the EPintermediate from
Equation 4 and integrating toyield:

The amplitude terms aredefined as follows:

APPENDIX

k1

product (ADP and Pi), respectively. Substrate binding was

assumed to be irreversible on the time scale of the quench
experiment in the sense that catalysis and product
release
were much faster than substrate release. This assumption is
reasonable and is, in fact, operationally defined by the ATP
quench/chaseexperiment because one onlyobserves ATP
that is converted to products during the chase
period. Preliminary evidence, based upon measurements of the rate of the
reverse reaction, the synthesis of ATP from ADP and phosphate, supports thisassumption.6
Scheme 1leads to thefollowing set of differential equations:

k3

E+S-ESeEP-E+P
k-z

(1)

where E,S , and P represent enzyme, substrate (ATP), and

ao =
011

k1k,[S]/[k1[SI(k2
+ k--8

+ k3) kzh]
+ ka) + kzk,]

(6)

kJd[kl[Sl(k+
, k-z

(7)

The values of a0 and a1determine the fractionsof enzyme as

EP and E, respectively, during the steady state.Accordingly,
the steady state turnover rate is given by v = k3a,,E0, such
that K,,, = kZk3/(k,(k, k-z + k,)) and V,,, = k,k,/(k, + kb2
+ kd.
The exponential rate constants are the roots of the quadratic equation:

h1.2

[ ( k l [ S ]+ kz

+ k-z + k3) & {(k,[S]+ kz + k-z +

- 4(k,[S](kz+ k-* + + kzk3)}21/2
k3)

(8)

k3)

The kinetics of ATP binding was measured by quenching

the reaction with an
excess of unlabeled ATP ona millisecond
time scale and then withacid after a period of time sufficient
to convertall tightly bound ATP to
products. The kineticsof
ATP hydrolysis (more precisely, the kineticsof the formation
of acid-labile phosphate) was measured by quenchingthe
reaction withacid on a millisecondtime scale (Johnson, 1983).
The ATP and Acid quench experiments measure the time
dependence of the following sums in unitsof product persite:
quench
Acid

[EP]/[Eo]+ [P]/[Eo]

(9)

ATP quench = [ESI/[&] + [EP]/[Eo]+ [Pl/[Eo]

= 1 -

[EI/[EoI + [Pl/IEol

K. M. Arndt and K. A. Johnson, unpublished observations.

10)

Downloaded from www.jbc.org by guest, on December 25, 2012

We have neglected effects due to multiple dynein heads.

There is some evidence, at low ATP concentrations, for the
kind of kinetic behavior expected for a three-headed dynein
molecule (Shimizu and Johnson,1983b), but furtherwork will
be required to establish the differences and functions of the
three heads. The observation that the ATP-induced
dissociation of the three-headed dynein molecule occurs as a single
exponential a t high ATP concentration might imply an inherent difference in the rate constantsfor the three heads or
may bedue topositive cooperativity in the ATP-binding
step.
It is significant that the cross-bridge cycle for the microtubule-dynein interaction is analogous to that described for
actomyosin (Lymn and Taylor, 1971; Taylor, 1979). In fact,
the kinetic constantsfor ATP binding, hydrolysis, and ATPinduced dissociation of the cross-bridge are withina factor of
2-3 (Taylor, 1979; Johnson and Porter, 1982b). The major
structural differences between dynein and myosin (Johnson
and Wall, 1983; Taylor, 1979) suggest that the similarities in
kinetic constants and pathway more
are likely to be afunction
of the general principles of catalysis rather than due to any
common ancestry. The ATP-induced
dissociation of the crossbridge is the critical step in establishing the pathway. Substrate-binding energy is presumably used to couple the ATP
hydrolysis reaction to thevectoral process (Jencks, 1980).
The major difference in rates is in the product-releasing
steps, where the rate observed for dynein exceeds that for
myosin by a factor of 100. This accounts for a corresponding
difference in the Michaelis constant. The slower turnover by
myosin may bea function of the required control mechanisms
where a muscle is in a relaxed state for long periods of time.
or flagellum startsbeating,it
Incontrast, onceacilium
continueswithoutinterruption.
As we learn more of the
similarities and differences between dynein and myosin, the
basic principles governing the conversion of chemical energy
to mechanical force production and regulation may become
more apparent.

Dynein Presteady State Phosphate Burst

13831

process was repeated and the final solution

was reached when
each constantgave a minimum in the sum square error, when
the other constants were held fixed a t their best estimate.
The sum square error was calculated as E( Yobs- Y,,,J2 for
each data set.
Confidence limitsforeachconstant
were evaluated by
constructing a confidence contour for each parameter (Draper
and Smith, 1981). The sum square errorwas calculated while
varying one parameter andholding the other three variables
fixed at their best estimate. The approximate
90% confidence
limits of each rate constantwere determined as the variation
Acid quenchamplitude = cy0 (1 - k3(X1 + X,)/X,X,)
(11)
in rate thatled to a 40-50% increase in the sum square error
(Draper and Smith,1981).
ATP quench amplitude = ao(k2 + k-, + k3
(12)
The best fit providesa unique solution. Since one must
- kMX1 + XZ)/XlXZ))/k2
simultaneously fit two data sets (ATP andAcid quench) and
The complete solution was programedinto a computer since each constant defines a particular aspect of the shape
by comgraphics subroutine and the best fit to thewas
data
evaluated of the curves, one cannot obtain two similar curves
visually and refined by an iterative nonlinearregression anal- pensating for a change in one constant by altering another.
ysis (Draper and Smith,
1981) based upon trial and error. TheMoreover, there isonly one global minimum in the sum square
steady state turnover rate was independently measured over error term for the data presented.
a period of several seconds immediately following each experREFERENCES
iment and was used in the fittingprocess to define the slope
Bagshaw, C. R., and Trentham, D. R. (1973) Biochem. J. 1 3 3 , 323of the line following the transient. In addition, the datawere
328
normalized to the amplitude of the presteady state ATPBlum, J. J., Hayes, A., Jamieson, G. A., Jr., and Vanaman, T. C.
binding transient in order to calculate the number of phos(1980) J. Cell Biol. 8 7 , 386-397
phates per site from the measured
value of phosphate concen- Brokaw, C. J. (1975) in Molecules and Cell Movement (Inoui, S., and
Stephens, R. E., eds) pp. 165-179, Raven Press, New York
tration. This calculation originally assumed an amplitude of
Brokaw, C. J., andBenedict, B. (1968) Arch. Biochem. Biophys. 125,
1 phosphate/site and was subsequently corrected to an am770-778
plitude of 0.85 phosphate/site according to Equation12 using Brokaw, C. J., and Gibbons, I. R. (1975) in Swimming and Flying in
the best estimates for the rate constants. This normalization Nature. (Wu, T. Y., Brennan, C., and Brokaw, C. J., eds) pp. 89factor was then used tocalculatetheapparent
molecular
126, Plenum Press, New York
Draper, N. R., and Smith, H.(1981) Applied Regression Analysis, 2nd
weight perATPbindingsitefromthemeasuredprotein
ed., John Wiley and Sons, Inc., New York
concentration.
Visual evaluation of the calculatedcurve provides an essen- Dyson, R. D., and Isenberg, I. (1971) Biochemistry 10, 3233-3241
Evans, J. A. (1982) Ph.D. thesis, University of Hawaii
tial criterion for judging the fit to the experimental data,
Gibbons, B. H., and Gibbons, I. R. (1972) J. Cell Bid. 54,75-97
especially during the initial stagesof the fitting process. The Gibbons, B. H., and Gibbons, I. R. (1974) J. Cell Biol. 63,970-985
majoradvantage of the visual evaluationisthatonecan
Gibbons, I. R. (1975) in Molecules and Cell Movement (Inoue, N. S.,
and Stephens, R. E., eds) pp. 207-232, Raven Press, New York
readily distinguish those aspects of a curve that are particularly sensitive toa given parameter. For example, the rate of Gibbons, I. R. (1981) J. Cell Biol. 9 1 , 107s-124s
ATP binding (kl[ATP]) primarily affects the shape
of the Gibbons, I. R., and Rowe, A. J . (1965) Science (Wash. D. C.) 1 4 9 ,
424-425
presteady state ATP-binding transient (ATP quench) and has
Goodenough, U. W., and Heuser, J. E. (1982) J. Cell Biol. 95, 798only secondary effects on the
fit to the linear phasefollowing
815
the transient or on the ATP hydrolysis curve. On the other Hackney, D. D., Stempel, K. E., and Boyer, P. D. (1980) Methods
Enzymol. 64, 60-83
hand, the net rate of ATP hydrolysis (kz + k-,) primarily
affects the shapeof the presteady state ATPhydrolysis tran- Huxley, H. E. (1969) Science (Wash. D. C.) 1 6 4 , 1356-1366
sient (Acid quench) and has little
effect on thefit to the linear Jencks, W. P. (1980) Adu. Enzymol. 51, 75-107
Johnson, K. A., and Porter, M. E. (1981) J. Cell Biol. 9 1 , 4 8 a
phase; the rate of ATP synthesis (k-,) primarily affects the Johnson,
K. A., and Porter, M. E. (1982a) Biophys. J. 37,345a
amplitude of the ATP hydrolysis transient. Only the rate of Johnson, K. A., and Porter, M. E. (198213) Cell Motility Supp. 1,101product release ( k B )has a marked effect on both the ATP
106
binding and the ATP
hydrolysis curves by affecting the slope Johnson, K. A., and Wall, J. S. (1983) J. Cell Bid. 9 6 , 669-678
Lymn, R. W., and Taylor, E. W. (1971) Biochemistry 10, 4617-4624
of the linear phasefollowing each transient.
Porter, M. E. (1982) Ph.D. thesis, University of Pennsylvania
The optimal method of obtaining the best fit to the data
Porter, M. E., and Johnson, K. A. (1983a) J. Biol. Chem. 258,6575took advantageof the uniqueeffects of each parameter on the 6581
two sets of experimental data. First, estimates for the rateof Porter, M. E., and Johnson, K. A. (1983h) J. Biol. Chem. 258,6582ATP binding (kl[ATP]) and the rate of product release ( k 3 )
6587
were obtained as the bestfit to the presteady state and linearSatir, P. (1968) J. Cell Biol. 39, 77-94
phases of the ATP-binding curve, respectively. At this stage Satir, P.,Wais-Steider, J., Lebduska, S., Nasr, A., and Avolio, J .
(1981) Cell Motility 1, 303-327
of the fitting process, initial guesses for the constantskZ and
Schendel, P. F., and Wells, R. D. (1973) J. Biol. Chem. 248, 8319k-2 could be used because these have only secondary effects
8321
on the fit to the ATP-binding curve. Next, estimates of the Shimizu, T. (1981) Biochemistry 20,4347-4354
values of k, and k-2 were obtained by fitting the net rate of Shimizu, T., and Johnson, K. A. (1983a) J . Biof. Chern. 258, 1383313840
ATP hydrolysis ( k , + k-*) to the presteady state phase and
fitting the rateof ATP synthesis(k-z) to the amplitudeof the Shimizu, T., and Johnson, K.A. (1983b) J. Biol. Chem. 2 5 8 , 1384113846
ATP hydrolysis (Acid quench) transient, using the values of
Shimizu, T., Kimura, I., Murofushi, H., and Sakai, H. (1979) FEBS
kl and k3 obtained in fitting the ATP-binding
curve. The best
Lett. 108, 215-218
estimates of k , and k+ were then used to refine the estimates Stein, L. A., Chock, P. B., and Eisenherg, E. (1981) Proc. Natl. Acad.
of k l and k3 in fitting the ATP-binding curve
again. The
Sci. U. S. A . 78, 1346-1350
The time dependence of [EP]/[Eo], [E]/[Eo], and [P]/[EO] are
defined by Equations 3, 4, and 5. Since the above equations
can most easily be solved in parts usinga microcomputer, we
have not made the substitutionsof Equations 3-8 into Equations 9 and 10for the final result.
The amplitudes of the Acid and ATP quench presteady
state transients are obtained by extrapolation of the linear
phase to t = 0. This can be defined algebraically by setting t
= 0 for the linear terms and t = 03 for the exponential terms
in the above equations to get thefollowing:

Downloaded from www.jbc.org by guest, on December 25, 2012

13832

Dynein Presteady State Phosphate Burst

Summers, K. E., and Gibbons, I. R. (1971) Proc. Natl. Acad. Sci.

U. S. A. 68,3092-3096
Takahashi, M., and Tonomura, Y. (1979) J. Biochem. (Tokyo) 8 6 ,
413-423
Taylor, E. W. (1977) Biochemistry 16,732-740
Taylor, E. W. (1979) Crit. Reu. Biochem. 6 , 103-164
Terashita, S., Kato,T.,Sato,
H., andTonomura, Y. (1983) J.
Biochem. (Tokyo)93,1575-1581
Tsukita, S., Tsukita, S., Usukura, J., and Ishikawa, H. (1983) J. Cell

Biol. 96,1480-1485
Warner, F. D., and Mitchell, D. R. (1980) Znt. Reu. Cytol. 6 6 , 1-43
Witman, G. B., and Minervini, N. M. (1982) in Proknryotic and
Eukaryotic Flagella (Amos, W. B., and Duckett, J. G., eds) pp. 203223, Cambridge University Press, Cambridge
Witman, G. B., Johnson, K. A., Pfister, K. K., and Wall, J. S. (1983)
J. Submicrosc. Cytol. 1 5 , 193-197
Zanetti, N. C., Mitchell, D. R., and Warner, F. D.(1979) J. Cell Biol.
80,573-588

Downloaded from www.jbc.org by guest, on December 25, 2012